LarpChat AI

LarpChat AI is a full-stack AI chat and image generation platform with tiered access. Users can chat with AI models and generate images, with different capabilities depending on whether they're a guest, free user, or Plus subscriber.

Founder and developer — sole engineer. Built the full stack: React frontend, Express backend, Supabase auth, OpenAI integration, rate limiting, usage tracking, and Vercel deployment pipeline.

• React + TypeScript — frontend SPA
• Express + TypeScript — backend server with serverless Vercel function entrypoints
• Supabase — email/password auth, user profiles, subscriptions, usage tracking
• OpenAI API — chat completions and image generation, server-side only
• Vercel — deployment with serverless API routes

• Guest: basic model access, strict IP-based rate limits, no account required
• Free: smarter model, higher daily limits, requires sign-in
• Plus: frontier model, highest limits, requires active subscription
• Plan enforcement happens entirely server-side — the client cannot spoof a tier

• Chat and image generation with tier-based model routing
• Supabase auth with session persistence across visits
• IP-based guest limits and user-based signed-in limits
• Daily usage tracking in a usage_events table
• Structured server-side logging for blocked requests and model usage
• PDF upload parsing (attachments gated to signed-in users)
• AI-generated chat titles using a lightweight model
• Public landing page, pricing page, and /app workspace

• OpenAI API key is server-only — never exposed to the browser
• Supabase service role key is server-only
• Row Level Security on all Supabase tables
• All plan enforcement and model selection happens on the backend
• Upload requests are minute-rate-limited server-side
• Structured logs capture abuse patterns without storing user content

Live at larpchatai.vercel.app. Active development — Stripe billing for Plus tier is the next planned milestone.